16
2110 Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation from Mexican Subduction Zone Earthquakes: The Attenuation Functions for Interplate and Inslab Events by T. Furumura and S. K. Singh Abstract The seismic waves from subduction zone earthquakes are significantly affected by the presence of 3D variation in crust and upper-mantle structure around the source area. These heterogeneous structures also profoundly modify the character of seismic waves as they propagate from the source area to regional distances. This is illustrated by studying shallow, interplate earthquakes along the Mexican subduc- tion zone, and deeper, inslab, normal-faulting earthquakes in the subducted Cocos plate beneath Mexican mainland. The strong-motion recordings of these earthquakes are used to evaluate the character of wave propagation along the path between the source region and Mexico City. We compare the wavefield from two large earth- quakes of different source type. During the shallow (H 17 km), interplate, 1995 Guerrero earthquake (M w 7.3), the Lg phase is the most prominent feature at re- gional distances of about 150 to a few hundred kilometers from the source. The presence of a lateral velocity gradient in the crust, caused by the subduction of the Cocos plate, enhances the Lg-wave amplitude, which is then amplified further in the Mexican volcanic belt by amplification in the low-velocity volcanic rocks. Both effects lead to very large ground motions along the path from the coast to the Mex- ican inland, in the frequency band from 0.2 to 4 Hz. However, for the deeper (H 40 km), inslab, normal-faulting, 1999 Oaxaca earthquake (M w 7.5), the amplitude of the Lg phase is too small to produce the abnormal wave propagation, and the direct S wave and its multiple SmS reflections between the free-surface and Moho show a simple attenuation with increasing distance. We compare these observations with numerical simulations of seismic-wave propagation using the Fourier spectral method. The results provide a key to the understanding of seismic-wave field gen- erated by shallow interplate and deeper inslab earthquakes in a realistic 3D hetero- geneous structure. Introduction Mexico suffers damage from interplate earthquakes that originate along the Pacific coast due to the subduction of oceanic plates below the continent. A recent and vivid ex- ample is the Michoacan earthquake of 1985 (M w 8.0), which devastated Mexico City, killing about 20,000 people (An- derson et al., 1986). The damage during the 1985 event was not surprising since Mexico City has repeatedly been struck by coastal earthquakes in the past. During this earthquake, severe ground shaking with a predominant frequency of about 0.3 Hz lasted for more than 3 min in the lake-bed zone of Mexico City located more than 300 km from the plate interface. The extensive damage in the city suggests anom- alous source, path, and/or site effects. Since 1985, significant progress has been made in quan- tifying and understanding the causes of abnormally large ground motions recorded in Mexico City from the Michoa- can earthquake in particular and interplate earthquakes in general. These include source studies (e.g., Campillo et al., 1989; Singh et al., 1988, 1990), studies on propagation ef- fects (e.g., Singh et al., 1988; Campillo et al., 1989; Ordaz and Singh, 1992; Furumura and Kennett, 1998) and local site amplification studies (e.g., Sa ´nchez-Sesma et al., 1988, 1993; Kawase and Aki, 1989; Cha ´vez-Garcı ´a and Bard, 1994; Cha ´vez-Garcia and Cuenca, 1996). The Mexican mainland is also vulnerable to large, nor- mal-faulting, inslab earthquakes that occur in the subducted ocean plate. There are many recent examples. The earth- quake of 15 January 1931 (M 7.8, H 40 km) caused severe destruction to the City of Oaxaca (Barrera, 1931; Singh et al., 1985); the earthquakes of 28 August 1973 (M w 7.0; H 82 km) and 24 October 1980 (M w 7.0; H 65 km) resulted in deaths and damage in the states of Veracruz, Puebla, and Oaxaca (e.g., Singh and Wyss, 1976; Yama- moto et al., 1984). The recent earthquake of 15 June 1999

Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

  • Upload
    others

  • View
    2

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2110

Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002

Regional Wave Propagation from Mexican Subduction Zone Earthquakes:

The Attenuation Functions for Interplate and Inslab Events

by T. Furumura and S. K. Singh

Abstract The seismic waves from subduction zone earthquakes are significantlyaffected by the presence of 3D variation in crust and upper-mantle structure aroundthe source area. These heterogeneous structures also profoundly modify the characterof seismic waves as they propagate from the source area to regional distances. Thisis illustrated by studying shallow, interplate earthquakes along the Mexican subduc-tion zone, and deeper, inslab, normal-faulting earthquakes in the subducted Cocosplate beneath Mexican mainland. The strong-motion recordings of these earthquakesare used to evaluate the character of wave propagation along the path between thesource region and Mexico City. We compare the wavefield from two large earth-quakes of different source type. During the shallow (H � 17 km), interplate, 1995Guerrero earthquake (Mw 7.3), the Lg phase is the most prominent feature at re-gional distances of about 150 to a few hundred kilometers from the source. Thepresence of a lateral velocity gradient in the crust, caused by the subduction of theCocos plate, enhances the Lg-wave amplitude, which is then amplified further inthe Mexican volcanic belt by amplification in the low-velocity volcanic rocks. Botheffects lead to very large ground motions along the path from the coast to the Mex-ican inland, in the frequency band from 0.2 to 4 Hz. However, for the deeper (H �40 km), inslab, normal-faulting, 1999 Oaxaca earthquake (Mw 7.5), the amplitude ofthe Lg phase is too small to produce the abnormal wave propagation, and the directS wave and its multiple SmS reflections between the free-surface and Moho show asimple attenuation with increasing distance. We compare these observations withnumerical simulations of seismic-wave propagation using the Fourier spectralmethod. The results provide a key to the understanding of seismic-wave field gen-erated by shallow interplate and deeper inslab earthquakes in a realistic 3D hetero-geneous structure.

Introduction

Mexico suffers damage from interplate earthquakes thatoriginate along the Pacific coast due to the subduction ofoceanic plates below the continent. A recent and vivid ex-ample is the Michoacan earthquake of 1985 (Mw 8.0), whichdevastated Mexico City, killing about 20,000 people (An-derson et al., 1986). The damage during the 1985 event wasnot surprising since Mexico City has repeatedly been struckby coastal earthquakes in the past. During this earthquake,severe ground shaking with a predominant frequency ofabout 0.3 Hz lasted for more than 3 min in the lake-bed zoneof Mexico City located more than 300 km from the plateinterface. The extensive damage in the city suggests anom-alous source, path, and/or site effects.

Since 1985, significant progress has been made in quan-tifying and understanding the causes of abnormally largeground motions recorded in Mexico City from the Michoa-can earthquake in particular and interplate earthquakes in

general. These include source studies (e.g., Campillo et al.,1989; Singh et al., 1988, 1990), studies on propagation ef-fects (e.g., Singh et al., 1988; Campillo et al., 1989; Ordazand Singh, 1992; Furumura and Kennett, 1998) and localsite amplification studies (e.g., Sanchez-Sesma et al., 1988,1993; Kawase and Aki, 1989; Chavez-Garcıa and Bard,1994; Chavez-Garcia and Cuenca, 1996).

The Mexican mainland is also vulnerable to large, nor-mal-faulting, inslab earthquakes that occur in the subductedocean plate. There are many recent examples. The earth-quake of 15 January 1931 (M 7.8, H � 40 km) caused severedestruction to the City of Oaxaca (Barrera, 1931; Singh etal., 1985); the earthquakes of 28 August 1973 (Mw 7.0;H � 82 km) and 24 October 1980 (Mw 7.0; H � 65 km)resulted in deaths and damage in the states of Veracruz,Puebla, and Oaxaca (e.g., Singh and Wyss, 1976; Yama-moto et al., 1984). The recent earthquake of 15 June 1999

Page 2: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2111

(Mw 7.0, H � 60 km) caused damage in the state of Puebla,especially to colonial structures in and near the city of Puebla(Singh et al., 1999). The 30 September 1999 Oaxaca earth-quake (Mw 7.5; H � 40 km) was damaging to the city ofOaxaca and many towns along the coast and between thecoast and the city of Oaxaca (Singh et al., 2000b). Suchsevere damage of low-rise building in the towns may havebeen caused by higher frequency ground shaking over 3 Hz.

There are only a few strong-motion recordings of inslabearthquakes that occurred prior to 1985. Although the seis-mic instrumentation has rapidly increased in Mexico begin-ning in 1985, significant inslab earthquake recordings onlybecame available beginning in 1994. For this reason, thesource studies of previous earthquakes were based on tele-seismic recordings (e.g., Singh and Wyss, 1976; Singh et al.,1985; Gonzalez-Ruiz, 1986). The estimation of ground mo-tion from inslab earthquakes was based on isoseismic inten-sities. For example, Singh et al. (1980) reported that theareas with modified Mercallli intensities (MMI) of V and VIfor inslab earthquakes were significantly larger than the cor-responding areas for interplate earthquakes of the samemagnitude. Since the seismic intensities are related to high-frequency ground motion of over 2 or 3 Hz, the larger MMIwould imply higher stress drop during inslab earthquakes ascompared to interplate earthquakes. Chavez and Castro(1988) investigated attenuation of the MMI for Mexicanearthquakes and reported that at epicentral distances of lessthan 200 km the attenuation was greater for interplate earth-quakes than for inslab ones. The opposite was the case atgreater distances.

In the past seven years, several significant inslab earth-quakes have occurred in the subducted Cocos plate belowMexico. These events were well recorded by the Mexicanbroadband seismograph network and the strong-motion ar-ray, permitting detailed source studies of some of theseevents (e.g., Mikumo et al., 1999; Quintanar et al., 1999;Hernandez et al., 2001; Iglesias et al., 2002). The data werealso used to derive empirical attenuation relations (Singh etal., 2000a; Garcia et al., 2001; Iglesias et al., 2002). Pachecoand Singh (1995) analyzed waveform from broadband seis-mographs and strong-motion network in Mexico, whichdemonstrated significant differences in the amplification ofseismic waves in the Valley of Mexico from interplate andinslab events. These studies capture, in a gross sense, thecharacteristics of the seismic-wave propagation from inslabsources but do not analyze the observed waveforms in termsof the depth of the source and its type acting in a complicated3D crust and upper-mantle structure.

The effect of the source type and its depth on the wavepropagation characteristics and the attenuation functions forboth the lower- (�0.5 Hz) and higher- (�3 Hz) frequencywave field is an important issue for understanding of thestrong ground motion during destructive earthquakes and forestimating the damage expected for future scenario earth-quakes. The aim of this study is to understand the wavepropagation and attenuation character of Mexican interplate

and inslab earthquakes and to investigate the effect of 3Dheterogeneous structure on the propagation of seismic waveat regional distances. In a previous article, Furumura andKennett (1998) examined the effect of the structure of thesubducted plate on the regional wave field as illustrated bythe observations from the interplate 1995 Guerrero earth-quake (Anderson et al., 1995) (Mw 7.3; H � 17 km). Inthis study we analyze the waveforms from strong groundmotion network and broadband seismographic stations of thenormal-faulting, inslab 1999 Oaxaca earthquake (Mw 7.5;H � 40 km) and compare them with those from the 1995earthquake. We will focus on the difference in the characterof the regional wave field from these different types ofsources and study the attenuation of seismic waves as a func-tion of frequency and source type.

To compliment the observations and gain further in-sights into the complicated seismic waveforms resultingfrom a 3D structure, we conduct numerical simulations ofseismic-wave propagation using the Fourier spectral method.In the process, we demonstrate the efficiency of 3D mod-eling that incorporates realistic crustal and upper-mantlestructure into the evaluation of strong ground motion fromfuture scenario earthquakes.

The Character of the Regional Seismic-WaveField in Mexico

Figure 1 displays the strong-motion accelerograph net-work of Mexico and the epicenters of the 1995 Guerreroearthquake (Mw 7.3) and 1999 Oaxaca earthquake (Mw 7.5),and Figure 2 illustrates the strong-motion records of radialand transverse component of velocity at rock sites along theprofile from the coast to Mexico City, covering a distancerange of 25 to 440 km. The velocity seismograms were ob-tained by integration of filtered accelerograms with a band-pass frequency of 0.02–20 Hz to reduce the instrumentalnoise. For display, each waveform is multiplied by the epi-central distance to compensate for the geometrical spread-ing, thus facilitating direct observation of abnormal waveamplification and attenuation as a function of epicentral dis-tance.

The records of the shallow, interplate, Guerrero earth-quake are characterized, in most part, by large-amplitude Lgwaves for epicentral distances beyond 150 km. These wavestravel with a wide range of group velocities, ranging between3.2 and 2.5 km/sec and arrive following a small-amplitudeSn phase.

It is well recognized that the Lg wave, as either the sumof higher-mode surface wave or of postcritical multiple SmSreflections between the free surface and Moho (see e.g., Ken-nett, 1985; Campillo, 1990), can propagate in the crustalwave guide at longer distances with less attenuation than theother S phases, such as direct S wave and the mantle Snwave. Therefore the Lg wave is usually the dominant featureat regional distances from 150 to over 1000 km in continen-tal structure. The Lg phase is usually the prominent arrival

Page 3: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2112 T. Furumura and S. K. Singh

16

17

18

19

20

101 100 99 98 97 96 95

COPLVIGA

OCTT

CHIL

MEZC

TNLP

IGUA

TEAC

CUER

MADR

TAMA

SMLC

OXLC

CHFL

RABO

YAIG

CUIG

Longitude (W)

La

titu

de

(N)

1995 GuerreroMw 7.3; H=17km

1999 OaxacaMw 7.5; H=40km

MVB

0

20

40

60

80

100 km

MexicoCity

0 100 km

Figure 1. Map of Mexican strong-motionnetwork and hypocenters of the 1995 Guerreroand 1999 Oaxaca earthquakes. Triangles,strong-motion stations; squares, broadband sta-tions. The depth of the Cocos plate is shownwith dashed lines. The boundary of the Mexi-can Volcanic Belt (MVB) and Mexico City areillustrated by a solid-line contour.

in the frequency band between 0.2 and 4 Hz (e.g., Herrmannand Kijko, 1983). The Lg phase is also greatly amplified asit propagates into the Mexican Volcanic Belt (MVB), at dis-tance of about 240 km from the epicenter, where low S-wavevelocity volcanic rocks underlie the surface with an approx-imate thickness of 2 km (e.g., Shapiro et al., 1997).

The amplitude of the Lg wave from the deeper, normal-faulting, inslab source of the 1999 Oaxaca earthquake (Fig.2b) is very small as compared to that of the shallow 1995Guerrero earthquake, because quite a bit of S-wave energyradiates into the crustal wave guide from the deeper event.The waveform from the inslab source displays a higher-frequency Sn phase and larger amplitude SmS reflections atsome distances. The large amplitude SmS phases are moredistinctly seen on the transverse component.

Compared with the 1995 Guerrero event, the length ofthe S-wave coda of the inslab earthquake is very short. Evenat MVB sites (e.g., stations YAIG, CUIG) the amplitude andlength of the SmS phases remain unchanged. Consequently,the amplitude of the S wave from this event decays simplywith increasing distances.

The empirical spectral attenuation of seismic wavesfrom interplate and inslab Mexican earthquakes has beenstudied by Ordaz and Singh (1992) and Garcia et al. (2001),respectively. The recordings at stations located in MVB wereexcluded from the analysis in both of these studies. Thespectral attenuation function, C(f, R), of the intense part ofthe ground motion can be described by

�pfR/V Q(f)sC(f, R) � e /G(R), (1)

where f is the frequency, R is the hypocentral distance, Q isthe anelastic attenuation coefficient, Vs is the shear-wavevelocity, and G(R) is the geometrical spreading term. Forinterplate earthquakes, Ordaz and Singh (1992) assumed Vs

� 3.5 km/sec, G(R) � R for R � 100 km, and G(R) �(100 R)1/2 for R � 100 km, and found Q(f ) � 273 f 0.66. Forinslab events, Garcia et al. (2001) assumed Vs � 4.5 km/secand G(R) � R for all distances and reported Q(f ) � 276f 0.57. The attenuation function is nearly the same for bothtypes of earthquakes in the distance (R � 500 km) and thefrequency range of interest in this study (0.3 � f � 4 Hz).

Figure 3 compares the spectral attenuation of the hori-zontal ground motion from the 1995 Guerrero and the 1999Oaxaca earthquakes at 0.5 and 4 Hz. An empirical attenua-tion function of Ordaz and Singh (1992) for shallow sub-duction earthquakes of Mexico and Garcia et al. (2001) isalso illustrated in the figure.

For the 1995 Guerrero earthquake the spectral attenua-tion at 0.5 Hz (Fig. 3a) shows a large discrepancy betweenthe observations and the empirical attenuation function atdistances beyond 200 km from the epicenter. As we haveseen in Figure 2, this difference arises, mostly, from thedominance of Lg phase in this frequency band and an ad-ditional amplification in the MVB. As mentioned previously,Ordaz and Singh (1992) excluded the data from MVB sitesin the derivation of the empirical attenuation function. Thusthe deviation of the observed data in the MVB from the em-pirical attenuation function for frequency of 0.5 Hz providesa measure seismic-wave amplification at these sites. Thecharacter of anomalous wave propagation in the MVB from

Page 4: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2113

(a) (b)

CO

PL

VIG

A

OC

TT

CH

IL

ME

ZC

TN

LP

IGU

A

TE

AC

CU

ER

CU

IG

MA

DR

TA

MA

SM

LC

OX

LC

CH

FL

RA

BO

YA

IG

CU

IG

100 200 300 400

Distance [km]

0

50

100

T-

Dis

tan

ce/8

.0[s

]

100 200 300 400 500

Distance [km]

0

50

100

T-

Dis

tan

ce/8

.0[s

]T T

(SmS)2

(SmS)3

SmS (PmP)2

(PmP)3

3.2

km/s2.5

PmP

Pn

Sn

(PmP)2

(PmP)3

(SmS)3

(SmS)2

SmS

Pn

Sn

Lg

PmP

100 200 300 400

Distance [km]

0

50

100

T-

Dis

tan

ce/8

.0[s

]

100 200 300 400 500

Distance [km]

0

50

100

T-

Dis

tan

ce/8

.0[s

]

R R

(SmS)3

(SmS)2

SmS

PmP

Pn

Sn

Lg

(PmP)2

(PmP)3

(SmS)3

(SmS)2

SmS

Pn

Sn

(PmP)2

(PmP)3

km/s2.5

3.2

PmP

Figure 2. Record sections of transverse-component (T) and radial-component (R)of ground velocity for the (a) 1995 Guerrero earthquake (H � 17 km) and (b) 1999Oaxaca earthquake (H � 40 km). The amplitude of each record is multiplied by theepicentral distance. The travel-time curves for a stratified reference model for Mexicoare superimposed on the records. Major phases are marked.

the shallow Mexican subduction zone earthquakes has beendiscussed by Ordaz and Singh (1992), Singh et al. (1995)and Cardenas et al. (1998) and is considered to be the causeof the large and long duration of ground motion at MexicoCity during the damaging earthquake (Cardenas et al., 1997;Shapiro et al., 1997; Furumura and Kennett, 1998; Iida,1999).

At the higher frequency of 4 Hz, the amplitude of bothinterplate and inslab earthquakes decays monochromaticallywith increasing distances, which are in good agreement withthe corresponding empirical attenuation functions, althoughthere is large scatter in the observation data of the interplateearthquake. The amplitude of the Lg phase at this frequency

is too small to produce the abnormal wave propagation aswe have seen in the 0.5-Hz wave field. We can find somelocalized ground amplitude at distances of about 100 and280 km from the epicenter caused by the first and secondarrivals of wide-angle SmS reflections.

2D Simulation of Seismic-Wave Propagation

To understand the character of the observed recordingsduring the 1995 Guerrero and the 1999 Oaxaca earthquakesand to examine the influence of the source depth and its typeand the structure of the Mexican subduction zone on thefeatures of the seismic-wave field, we conduct 2D numerical

Page 5: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2114 T. Furumura and S. K. Singh

0.10.1

1

10

100 200 300 400

0.10.1

1

10

100 200 300 400

0.10.1

1

10

100 200 300 400 500

0.10.1

1

10

100 200 300 400 500Distance [km] Distance [km]

Distance [km]

Am

plitu

de [c

m/s

]

Distance [km]

Am

plitu

de [c

m/s

]

Am

plitu

de [c

m/s

]A

mpl

itude

[cm

/s]

4 Hz

0.5 Hz

(a) (b)

4 Hz

0.5 Hz

Figure 3. Spectral attenuation of horizontalground motion as a function of hypocentraldistance for the (a) 1995 Guerrero earthquake(H � 17 km) and (b) 1999 Oaxaca earthquake(H � 40 km) at (top) 0.5 Hz and at (bottom)4 Hz. The continuous lines represent the em-pirical spectral attenuation function for Swaves (see text).

Table 1Model Parameters Used in the 3D Simulation

Vp

(km/sec)Vs

(km/sec)q

(t/m3) QThickness

(km)

Mexican MainlandLayer 1 (MVB) 4.0 2.0 2.0 150 0.0–2.5Layer 2 5.2–5.3 3.0–3.1 2.2–2.3 200 0.6–8.0Upper crust 5.5–5.8 3.2–3.4 2.4–2.5 400 4.9–12.6Lower crust 6.4–7.1 3.7–4.1 2.7–2.9 600 11.5–29.8Upper mantle 8.2–8.4 4.7–4.8 3.2–3.2 800 —

Cocos PlateOceanic crust 5.0–7.0 2.8–3.9 2.1–2.8 100–150 3.0Oceanic basalt 6.8–7.1 3.8–3.9 2.8–2.9 150–200 3.0Oceanic mantle 8.2–8.6 4.7–4.9 3.2–3.3 1000 24.0

simulations using the Fourier spectral method (e.g., Kosloffet al., 1984; Furumura and Takenaka, 1996).

In the simulations, we use a crustal and upper-mantlestructure of the Mexican subduction zone that includes theCocos plate descending beneath Mexican mainland with anaverage dip angle of 10�. The model is adopted from theworks of Valdes et al. (1986) and Valdes and Meyer (1996),which are based on an analysis of reflection experiment andgravity data. The subducted plate consists of a two-layer,6-km-thick oceanic crust with relatively low-velocity andhigh-attenuation material, overlaying a higher-velocity andlow-attenuation oceanic mantle. The thickness of the litho-sphere is 30 km (Table 1). We have included surface topog-raphy with a maximum height of 2 km above the sea leveland a 2-km-thick superficial layer to represent the MVB withslow shear-wave velocity (Vs � 2.0 km/sec) and a qualityfactor for S wave of 150.

The model covers an area of 512 km in length and 128km in depth, which is discretized at a uniform grid spacingof 0.25 km, with a 20-grid points buffer zone of Cerjan etal. (1985) in the surrounding area to minimize artificial re-flections at the boundaries.

This 2D simulation is valid for frequencies up to 4 Hz,which is four times higher than the previous experiments(Furumura and Kennett, 1998).

We have used a low-angle thrust-fault source at a depthof 17 km to represent the 1995 Guerrero earthquake sourceand a normal-faulting source at a depth of 40 km for the1999 Oaxaca earthquake. The source time function is apseudodelta function (Herrmann, 1979), which imparts seis-mic waves with a maximum frequency of 4 Hz.

Interplate Thrust-Faulting Source

Wave-field snapshots of P and SV waves and syntheticseismograms of radial-component ground motion for the in-terplate earthquakes are displayed in Figure 4a.

In the upper snapshot frame (6 sec) we see the radiationof P and S waves from the source on the boundary betweenthe subducted plate and the continental lower crust. The Swave, with an incidence angle greater than 35� at the freesurface, is totally reflected back into the crust (24 sec). Theshallow-dipping subducted Cocos plate also acts as an effi-cient reflector for the S wave, which leads to clear multipleS-wave reverberations between the free surface and the dip-ping plate. By 54 sec, a well-developed Lg-wave train, asthe sum of postcritical SmS reflections between the free sur-face and Moho (or the subducting plate), is clearly seen in

Page 6: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2115

100. 200. 300. 400.

0.

25.

50.

75.

100.

T-

Dis

tan

ce/8

.0[s

]

100. 200. 300. 400.

0.

25.

50.

75.

100.

Distance [km]

T-

Dis

tan

ce/8

.0[s

](b)(a)

(SmS)3

(SmS)2

(SmS)

Lg

Sn

Pn

PmP

(PmP)2

(PmP)3(SmS)2

(SmS)3

SmS

Sn

Pn

(PmP)3

(PmP)2

PmP

2.5km/s

3.2km/s

R R

54 s

96 s

24 s

6 s

Lg

Pn

SnLg

P

S Moho

MVB

50.

De

pth

[km

]

0. 200. 300. 400.

0.

Distance [km]

100.

Sn

Mantle

Crust

Distance [km]

6 s

21 s

42 s

69 s

Pn

Sn

Sn

Pn

sP

(SmS)2

P

SMoho

MVB

50.

0. 200. 300. 400.

0.

Distance [km]

100.

SmS

SmS

Mantle

Crust

Plate Plate

Figure 4. Synthetic velocity seismograms for the radial component of ground mo-tion and snapshots of seismic-wave field calculated from 2D numerical modeling forthe (a) interplate thrust-faulting event (H � 17 km) and (b) inslab normal-faultingevent (H � 40 km). The P wave is shown in red, and the S wave is shown in green.The travel-time curves for a stratified reference model are superimposed on the recordsection, and major phases are marked.

the crust. We can observe trapped S-wave energy propagat-ing in the crustal wave guide, which travels long distanceswithout attenuation (96 sec). We can also note multiple S-wave reflections between the free-surface and midcrustal in-terface (i.e., the Conrad discontinuity), so that the sum ofthese reflections produces a dense V-shape wave train prop-agating in the crust (96 sec).

Synthetic seismograms of the radial-component groundvelocity for the shallow interplate event are displayed in Fig-ure 4a, where regular pattern of multiple SmS and Conradreflections are clearly seen at distances over 150 km. The

Lg-wave train arrives with a group velocity between 3.2 and2.5 km/sec, in good agreement with the observations (Fig.2a). The amplitude of the Sn head wave is too weak to dis-cern in the seismograms, which also agrees with the obser-vations. The weak mantle Sn phase is in consequence ofsmall fraction of S-wave energy leaking from the crust tothe mantle.

Inslab Normal-Faulting Source

Snapshots of seismic-wave field and synthetic seismo-grams for an inslab earthquake, displayed in Figure 4b, show

Page 7: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2116 T. Furumura and S. K. Singh

a markedly different pattern of seismic-wave field as com-pared to that for the shallow source.

In the first frame of the snapshots (6 sec), we see theradiation of the S wave from the source inside the plate. TheS wave impinges upon the free surface with a small inci-dence angle of about 10� at a distance around 30 km fromthe epicenter and causes large ground oscillations there. Thesmall angle of incidence of S wave at the free surface pro-duces very large S to P conversion, which dramaticallyweakens the S-wave reflections (21 sec). The S phase is fur-ther weakened on reflection from the subducting plate andthe Moho, most of the energy leaking into the mantle (42 sec,69 sec). Thus, the reverberating SmS phase in the crust de-creases dramatically in amplitude with propagation throughthe leaking of energy into the mantle.

The change in the character of regional seismic-wavefield as a result of increasing source depth is also clearlyseen in the synthetic seismograms (Fig. 4b). For instance,Sn arrivals are large in the distance range of 120–400 km,with increasing relative amplitudes at larger distances. Thereare clear multiple SmS reflections in the seismograms witha 15-sec interval between phases. Reflections at the Conradare very weak for the 40-km-depth source. The amplitude ofseismic waves decays in a relatively simple manner withincreasing epicentral distance.

Decay Function for Interplate and Inslab Earthquakes

Figure 5 displays decay functions of horizontal motionderived from the 2D simulations of the thrust-fault, interplateearthquake and a normal-fault, inslab event at frequencies0.5 and 4 Hz. The empirical attenuation functions of Ordazand Singh (1982) and Garcia et al. (2001) for interplate andinslab earthquakes, respectively, are superimposed on thefigure.

As we have seen in the observed seismograms, there issignificant difference in the attenuation for the two types ofearthquakes, especially at 0.5 Hz where the Lg wave domi-nates. For shallow, interplate earthquakes, the peak groundvelocity starts increasing at distances beyond 200 km fromthe epicenter. The peak amplitude at 320 km is roughly threetimes greater than the predicted amplitude from the empiri-cal attenuation function. This observation is in agreementwith the results of Ordaz and Singh (1992). The simulatedwaveforms from the normal-faulting inslab earthquake showlarger amplitude at shorter distance (�120 km) and slightlylarger attenuation than average at longer propagation dis-tances at 0.5 Hz.

At the higher frequency of 4 Hz, the amplitudes for bothtypes of earthquakes decay in a simpler manner with increas-ing distance than at 0.5 Hz. The simulation result agrees withthe observation reasonably well, although there is large scat-ter in the observed high-frequency (4 Hz) wave field fromthe interplate earthquake. Localized amplification at dis-tances of 100 and 280 km from epicenter caused by an ar-rival of wide-angle SmS reflections is also simulated quitewell.

Figure 5 provides a possible explanation for the obser-vation of Chavez and Castro (1988). Because of the bumpin the attenuation curve for interplate earthquakes beginningabout 200 km caused by the amplification in the MVB andlarge amplitude SmS phases, the attenuation rate changesaround this distance. Thus, the attenuation from interplateearthquakes appears slower beyond about 200 km. This isnot the case for inslab earthquakes.

Figure 6 compares the Fourier spectra computed fromsynthetic seismograms at epicentral distances of 40, 120, and320 km. At 40 km, the Fourier amplitude from the inslabearthquake is greater than from the interplate event. How-ever, as the epicentral distance increases, the Fourier ampli-tudes of the shallow interplate earthquake increase graduallyin the lower frequency band of 0.3–4 Hz. At the epicentraldistance of 320 km, the amplitude from the interplate earth-quake is almost 10 times greater than that for the inslabearthquake in this frequency band.

This figure demonstrates that the heterogeneous crustalstructure of the Mexican mainland, which is characterizedby subducted Cocos plate and shallow structure of the MVB,acts a good propagator for S waves in the frequency band of0.3–4 Hz for the shallow interplate earthquakes.

3D Simulation of Mexican Earthquakes

We conducted 3D simulation of seismic-wave propa-gation from the two Mexican earthquakes to compliment theresults of the 2D modeling and to study the character of wavepropagation in realistic 3D heterogeneous structure of Mex-ican mainland.

We employ a Fourier spectral method (PSM) finite-difference method (FDM) hybrid simulation code (Furumuraet al., 2000, 2002) for large scale 3D simulation using par-allel computers. This combines the Fourier spectral method(PSM) in the calculation of seismic-wave field in the hori-zontal (x, y) coordinates with the fourth-order finite-differencemethod (FDM) in the vertical (z) coordinate. The accuratePSM calculation allows large-grid discretization in the hor-izontal direction, and the localized FDM calculation mini-mizes the demand for the interprocessor communication forthe parallel computing as compared to the conventional par-allel PSM (e.g., Reshef et al., 1988; Furumura et al., 1998;Furumura and Koketsu, 2000). Therefore, the parallel PSM/FDM hybrid method offers high parallel performance forlarge-scale 3D simulation using a large number of proces-sors.

The simulation model is 512 km by 512 km by 104 km,which is discretized with a uniform grid size of 1 km in thehorizontal direction and 0.5 km in the vertical direction. Thesubsurface structure is based on the Vp and the density (q)structure derived by refraction experiments and gravity data(Valdes et al., 1986; Valdes and-Meyer, 1996), the Vs struc-ture estimated from the surface wave (Campillo et al., 1996;Shapiro et al., 1997), the quality factor (Q) from spectralattenuation of S and Lg waves (Castro et al., 1994; Cardenas

Page 8: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2117

0.10.1

1

10

100 200 300 400

0.10.1

1

10

100 200 300 400

0.10.1

1

10

100 200 300 400

0.10.1

1

10

100 200 300 400

(a) Interplate source (H=17 km) (b) Inslab source (H=40 km)

Rel

ativ

e A

mpl

itude

Distance [km]

Rel

ativ

e A

mpl

itude

Distance [km]

Rel

ativ

e A

mpl

itude

Distance [km]

Rel

ativ

e A

mpl

itude

Distance [km]

4 Hz

0.5 Hz

4 Hz

0.5 Hz

Figure 5. Attenuation of simulated peak amplitude of ground velocity as a functionof epicentral distance and frequency from the modeling of (a) interplate thrust-faultingevent (H � 17 km) and (b) inslab normal-faulting event (H � 40 km) at frequencyof (top) 0.5 Hz and (bottom) 4 Hz. The amplitude is normalized to the same value at100 km.

et al., 1997; Ottemoller et al., 2002), and the depth of sub-ducted plate from hypocentral data (Pardo and Suarez, 1995;Kostoglodov et al., 1996). The shape of Moho is derivedfrom the surface topography by assuming 80% isostaticcompensation and that the Moho surface is smooth. Themodel is shown in Figure 7.

We assigned lower velocity (Vp � 4.4 km/sec Vs � 2.0km/sec) and slightly larger attenuation (Q � 150) to the2-km-thick volcanic rocks of the MVB. We assigned a rela-tively faster velocity for the water layer of Vp � 2.2 km/secand Vs � 0 km/sec and Q � 200, because the velocity ofwater layer is too slow to include in the simulation model.The model includes an accretionary prism along the trench,which has low velocity (Vp � 4.4 km/sec Vs � 2.2 km/sec),a high attenuation (Qs � 50), a width of about 40 km, anda maximum thickness of about 4 km. Surface topography isnot included in the simulation model.

The minimum S-wave velocity of the superficial layerin the 3D model is 2.0 km/sec, so that with sampling of twogrid points per shortest wavelength in the horizontal coor-dinate and four grid points in the vertical coordinate, thePSM/FDM hybrid method can simulate seismic-wave prop-agation for frequencies below 1 Hz. The 3D simulation tooka memory of 8 Gbyte and a wall-clock time of 9 hr using32 processors of a Fujitsu VPP700E vectorized parallel com-puter for calculating wave propagation for 210 sec through7000 time-step computations.

The 1995 Guerrero Earthquake

The fault-slip model for the 1995 Guerrero earthquakewe use here is derived from an inversion of strong-motiondata (Courboulex et al., 1997). The slip distribution showstwo branches aligned toward the south and east, and maxi-mum slip of over 3 m occurs 10 km south of the hypocenter

Page 9: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2118 T. Furumura and S. K. Singh

Inslab source (H=40 km)

Interplate source (H=17 km)

10.

1.

0.1

Fou

rier

Am

plitu

de

0.1 1.Frequency [Hz]

(a) (b) (c)

0.3 3.

10.

1.

0.1

10.

1.

0.1

0.1 1.Frequency [Hz]

0.3 3. 0.1 1.Frequency [Hz]

0.3 3.

Figure 6. Comparison of Fourier amplitude spectra of simulated waveforms for theshallow (H � 17 km) thrust-faulting event (solid line) and deeper (H � 40 km) inslabnormal-faulting event (dashed line) at epicentral distance of (a) 40 km, (b) 120 km,and (c) 320 km.

100 km

0

20

40

60

80

45

4035

15

40km

40

Moho

Plate

100.

300.

500.

Y [km]

100.

300.

500.

X[km]

50.

100.Z[k

m]

Coast

MVB

Moho

Plate

15

16

17

18

19

20

101 100 99 98 97 96 95

MexicoCity

MVB

b

b’

Longitude (W)

La

titu

de

(N)

100 km

0.0

1.0

2.0

3.0

Slip [m]

15.

20.

10.

Dep

th[k

m]

50.

30.

40.

Dep

th[k

m]

Distance along strike [km]

20

4060

80

2040

1995 GuerreroEarthquake

1999 OaxacaEarthquake

0 0

a

a’

Figure 7. Structural model for Mexico showing configuration of the Moho discon-tinuity and the Cocos plate. Source-slip model for the 1995 Guerrero (Courboulex etal., 1997) and the 1999 Oaxaca earthquakes (Hernandez et al., 2001) are also displayed.

Page 10: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2119

at the southern branch. The ruptures propagate bilaterallyfrom the center with a rupture velocity of 2.5 km/sec (Cour-boulex et al., 1997).

To allow smooth fault rupture propagation, the faultplane is divided into 450 patches of size 2.5 km � 2.5 kmby linear interpolation, and a double-couple point source isassigned to each subfault. The source-time history of eachpoint source is a pseudodelta function (Herrmann, 1979)with a dominant period of 1 sec. The total seismic momentis M0 � 1.31 � 1020 N m.

Figure 8a shows a set of snapshots of the horizontalground velocity. In the first frame (21 sec) a large amplitudeS wave radiates from the source, and in the second frame(44 sec), an Sn wave propagates to a distance of about 200km from the epicenter. The sequence of multiple SmS re-flections between the free surface and Moho produce largeamplitude Lg phase; the duration of Lg-wave train increasesas it propagates inland to the Mexican mainland (67 sec).The Lg phase tends to reduce its propagation speed on en-tering the MVB and acquires a large amplitude through theamplification effect of the low-velocity superficial layer (99sec).

Synthetic velocity seismograms for the radial and trans-verse components along the station profile of aa�, which liesclose to the recording stations shown in Figure 1, are dis-played in Figure 9a. In Figure 10 we display the distributionof the peak ground acceleration (PGA) of the horizontal mo-tion. Larger values of PGA from the source toward inlandare clearly seen in Figure 10. These results, in large part,come from the propagation of the Lg wave and its localamplification in the MVB. The pattern of anomalous inlandamplification of PGA agrees well with the observed PGA,although the coverage of strong-motion stations to the south-east of Mexico City is not dense enough to produce a de-tailed distribution of PGA during the earthquake. We com-pare the simulated and observed PGA values with thatexpected from an empirical function of Ordaz and Singh(1989) for interplate Mexican earthquakes in Figure 11. Thisfunction is derived from extensive hard-rock site strong-motion recordings (Ordaz and Singh, 1989). The simulatedPGA values are in good agreement with the correspondingempirical function for hard-rock stations. Larger PGA valuesat MVB stations also agree well with observations.

The 1999 Oaxaca Earthquake

Figure 8b shows the snapshots of horizontal velocityfrom the 3D simulation of the 1999 Oaxaca earthquake, andFigure 9b illustrates the corresponding synthetic seismo-grams along a profile of bb� in Figure 7, close to the stationprofile in Figure 2.

We used the fault rupture model of Hernandez et al.(2001) derived from an inversion using near-source strong-motion records. The slip distribution on the fault shows twoareas of 2.5-m peak slip located about 20 km and 40 kmnorthwest of the hypocenter at a depth of about 45 km.

In this calculation the fault rupture is approximated by

630 double-couple point sources arranged on the fault platewith an interval of 2.7 km, and the rupture propagates almostunilaterally from the epicenter to the northwest with a rup-ture velocity of about 2.7–3.0 km/sec (Hernandez et al.,2001). Total moment for this earthquake is M0 � 1.7 �1020 N m.

The snapshot of the seismic-wave field from the deepersource of the 1999 Oaxaca earthquake shows simple char-acteristics as compared with that from the shallow, thrust-faulting 1995 Guerrero earthquake (Fig. 8). The simulatedPGA distribution shows a relatively simple decay with dis-tance with multiple localized zones of high amplitudes about200 and 400 km from the epicenter. These zones are pro-duced by arrival of multiple SmS reflections (Fig. 10b). Theamplification at the MVB is insignificant for the vertical in-cidence of these S phases from the deeper inslab source.

We compare the simulated PGA values with that ex-pected from an empirical function of Garcia et al. (2001) forMexican inslab earthquakes in Figure 11. The attenuationcharacter and PGA distribution from the 3D simulation forthe 1999 Oaxaca earthquake agrees well with the observa-tions, but the simulated PGA is about half of the observedvalues. This is because the higher-frequency waves, whichare more important from the deep inslab earthquakes (e.g.,Singh et al. 2000a), are not included in the present 3D simu-lation.

In Figure 9 we display the simulated record sections ofhorizontal component seismograms along the station profileof bb� in Figure 7. The SmS reflections from this source havelarger amplitudes on the transverse component, which alsoagrees well with the observations.

Discussion

It has long been recognized that the Valley of Mexicois vulnerable to shallow coastal earthquakes that occur onthe interface of the subducting Cocos plate, even though thevalley is locating over 300 km from the coast. The damagefrom interplate earthquakes may, in part, be due to the dom-inance of the Lg wave propagating in the crustal wave guide.The subhorizontal geometry of the subducted Cocos plateacts as perfect reflector for S waves from interplate earth-quakes. This can greatly enhance the Lg amplitude in thefrequency range between 0.2 and 4 Hz. In volcanic areas,the near-surface structure of low velocity also increases theamplitude and the duration of the Lg wave. Therefore thecrustal structure of Mexican mainland acts a good propa-gator of S-wave energy in the frequency band of about 0.2–4 Hz. This frequency band is closely related to the resonantfrequency of the clay layers in the lake-bed zone of MexicoCity (0.3 to 0.7 Hz) (Singh et al., 1988), so the Lg waveshould acquire an additional amplification there (Campilloet al., 1988; Furumura and Kennett, 1998).

However, when the event occurs inside the plate, the Swave radiated from the deeper source and its SmS multiplesin the crust impinges upon the surface with almost vertical

Page 11: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2120 T. Furumura and S. K. Singh

Figure 8. Snapshots of horizontal ground velocity from 3D simulation of seismic-wave propagation from (a) the 1995 Guerrero earthquake and (b) the 1999 Oaxacaearthquake. The epicenters are marked by crosses. Time from the origin is shown inthe bottom right corner.

Page 12: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2121

(a)

(b)

b-

-a'

-b'

a-

10

0.

20

0.

30

0.

40

0.

0.

25

.

50

.

75

.

10

0.

Dis

tan

ce[k

m]

T-Distance/8.0[s]

10

0.

20

0.

30

0.

40

0.

50

0.

0.

25

.

50

.

75

.

10

0.

Dis

tan

ce[k

m]

T-Distance/8.0[s]

Sn

Pn

(Pm

P)3

(Pm

P)2

Pm

P

Sm

S

(Sm

S)2

(Sm

S)3

(Pm

P)3

(Pm

P)2

Pm

P

(Sm

S)3

(Sm

S)2

SmS

Pn

Sn

3.2

Lg

TT

3.2

3.2

2.5 km/s

2.5 km/s

10

0.

20

0.

30

0.

40

0.

50

0.

0.

25

.

50

.

75

.

10

0.

Dis

tan

ce[k

m]

T-Distance/8.0[s]

10

0.

20

0.

30

0.

40

0.

0.

25

.

50

.

75

.

10

0.

Dis

tan

ce[k

m]

T-Distance/8.0[s]

Sn

Pn

(Pm

P)3

(Pm

P)2

Pm

P

SmS

(Sm

S)2

(Sm

S)3

(Pm

P)3

(Pm

P)2

Pm

P

(Sm

S)3

(Sm

S)2

SmS

Pn

Sn

3.2

Lg

RR

3.2

3.2

2.5 km/s

2.5 km/s

b-

-a'

-b'

a-

Figu

re9.

Rad

ial

and

tran

sver

seco

mpo

nent

reco

rdse

ctio

nsfr

om3D

sim

ulat

ion

ofw

ave

prop

agat

ion

from

(a)

the

1995

Gue

rrer

oea

rthq

uake

alon

gaa

�in

Figu

re7

and

(b)

the

1999

Oax

aca

eart

hqua

keal

ong

bb�

inFi

gure

7.T

hetr

avel

-tim

ecu

rves

for

ast

ratifi

edre

fere

nce

mod

elar

esu

peri

mpo

sed

onth

ere

cord

s.

Page 13: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2122 T. Furumura and S. K. Singh

132.7

72.5

22.6

14.1

18.2

13.616.4

11.7

5.811.1

11.2

60.577.0

299.1

245.66.0

27.3

5.7

3.3

224.8

103.4332.7

8.0 180.5

20

50

10

5

200300

100

20

100

PGA[cm/s/s]

PGA[cm/s/s]

(b)

3.3

13.1

12.88.1

12.8

70.4

102.369.7

7.0

8.8 98.0

27.3

10.6

41.270.812.4

41.7 52.6

PGA[cm/s/s]

0 100 km

(a)

100

20

10

PGA[cm/s/s]

100

50

20

10

Cal.

Obs.

Cal.

Obs.

50

102.6

50

5

10

0 100 km

MVB MVB

MVBMVB

200

0 100 km 0 100 km

Figure 10. Peak horizontal ground acceleration. (a) The 1995 Guerrero earthquake(Mw 7.3) (b) the 1999 Oaxaca earthquake (Mw 7.5). (Top) 3D simulation; (bottom)observations.

incidence angle. This is a major loss of S-wave energy byconversion to a P wave at the free surface. The amplificationdue to the shallow structure of the MVB is also insignificantfor near-vertical incidence of the S wave. Since the wave-form from the inslab earthquake is characterized by the Snwave and some isolated weak SmS multiples, the durationof ground motion is much shorter than during interplateevents with prolonged Lg-wave trains. The result is that theS wave from inslab earthquakes decays simply with increas-ing distances for all frequencies. This suggests that the dam-

age area from inslab earthquakes should be smaller than thatfor interplate earthquakes, unless the stress drops during for-mer events are much higher than for the latter ones. Recentstudies on comparing seismic intensities during interplateand inslab earthquakes in Mexico suggest that the groundmotions and hence the stress drop involved in the formertype of events is higher than in later one (Singh et al., 2000a;Garcia et al., 2001; Iglesias et al., 2001).

Numerical 3D simulation of seismic-wave propagationusing a realistic 3D model for the structure of Mexico and

Page 14: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2123

0.10.1

1

10

100

1000

10 100100 1000

0.10.1

1

10

100

1000

10 100100 1000

0.10.1

1

10

100

1000

10 100100 1000Distance [km] Distance [km]

PG

A [

cm/s

/s]

(a) (b)

PG

A

[cm

/s/s

]

Mw 7.3 Mw 7.5

Attenuation function for Interplate events:

log PGA = 1.760 + 0.300 Mw - log R - 0.0031 R

Ordaz and Singh (1992)Attenuation function for Inslab events:

log PGA = -0.148 + 0.623 Mw - log R - 0.0032 R

Garcia et al. (2001)

Rock station (Cal.)

MVB stations (Cal.)

- (Obs.)

- (Obs.)

Rock station (Cal.)

MVB stations (Cal.)

- (Obs.)

- (Obs.)

Figure 11. Comparison of observed and calculated peak horizontal ground accel-eration (PGA) as a function of focal distance. (a) The 1995 Guerrero earthquake; (b)the 1999 Oaxaca earthquake. The PGA values inside the MVB are shown in white, andthose outside the MVB are in black. Also shown are expected PGA value, based onregression analysis, for a shallow interplate Mw 7.3 earthquake and a Mw 7.5 inslabearthquake.

adequate source models provides a reasonable representationof the observed ground motion from Mexican subductionzone earthquakes for both a shallow interplate event and adeep inslab event. The model can potentially be used to in-vestigate the pattern of ground motion expected for futureevents, such as at current seismic gaps, although even usingcurrent powerful computers it may still be difficult to pro-vide an adequate resolution of the high-frequency wave fieldfrom inslab earthquakes because of the limited knowledgeof heterogeneous crust and upper-mantle structure and thesource-slip models during the earthquakes.

One of the important issues of strong-motion seismol-ogy is the evaluation of the damage potential in urban centersand its reduction during future scenario earthquakes. Theresults of this study suggest that a close link between sourcestudies, strong-motion observations, detailed knowledge ofcrustal structure, and numerical simulation technology is in-dispensable to make substantial advances in this direction.

Acknowledgments

The collaboration of the authors has been supported by a grant-in-aidfrom the Japan Society for the Promotion of Science for foreign studies.The research was also partly supported by the DGAPA, UNAM projectsIN109598 and IN111601. We are grateful to T. Mikumo for fruitful dis-cussions and for his constructive remarks. The computation was conductedat the Earthquake Information Center, University of Tokyo, and at the In-

stitute of Physics and Chemical Research, Japan (RIKEN). Constructivecomments from the critical reading of the manuscript by Dr. R. B. Herr-mann and an anonymous reviewer have been very helpful for the revisionof the article.

References

Anderson, J. G., P. Bodin, J. N. Brune, J. Prince, S. K. Singh, R. Quaas,and M. Onate (1986). Strong ground motion from the Michoacan,Mexico, earthquake, Science 233, 1043–1049.

Anderson, J. G., R. Ruaas, S. K. Singh, J. M. Espionsa, A. Jiminez, J.Lermo, J. Cuenca, F. J. Sanchez-Sesma, R. Meli, M. Ordaz, S. Al-cocer, B. Lopez, L. Alcantara, E. Mena, and C. Javier (1995). TheCopla, Guerrero, Mexico earthquake of September 14, 1995 (Mw �

7.4): a preliminary report, Seism. Res. Lett. 66, 11–39.Barrera, T. (1931). El temblor del 14 de enero de 1931, Report, Instituto

de Geologia UNAM, Mexico City, Mexico.Campillo, M. (1990). Propagation and attenuation characteristics of the

crustal phase Lg, Pure Appl. Geophys. 132, 1–9.Campillo, M., P.-Y. Bard, F. Nicollin, and F. Sanchez-Sesma (1988). The

Mexico earthquake of September 19, 1985 the incident wavefield inMexico City during the great Michoacan earthquake and its interpre-tation with the deep basin, Earthquake Spectra 4, 591–608.

Campillo, M., J. C. Gariel, K. Aki and F. J. Sanchez-Sesma (1989). De-structive strong ground motion in Mexico city: Source, path and siteeffects during great 1985 Michoacan Earthquake, Bull. Seism. Soc.Am. 79, 1718–1735.

Campillo, M., S. K. Singh, N. Shapiro, J. Pacheco, R. B. Herrmann (1996).Crustal structure south of the Mexican volcanic belt, based on groupvelocity dispersion, Geofısica Int. 35, 361–370.

Cardenas, M., F. J. Chaez-Garıa, and A. Gusev (1997). Regeonal amplifi-

Page 15: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

2124 T. Furumura and S. K. Singh

cation of ground motion in central Mexico. Results from coda-lengthmagnitude data and preliminary modeling, J. Seismology 1, 341–355.

Cardenas, M., F. Nunez-Cornu, J. Lermo, D. Cordoba, and A. Gonzalez(1998). Seismic energy attenuation in the region between the coast ofGuerrero and Mexico City: differences between paths along perpen-dicular to the coast, Phys. Earth. Planet. Int. 105, 47–57.

Castro, R. R., L. Munguıa, C. J. Rebollar, and J. G. Acosta (1994). Acomparative analysis of the quality factor Q for the regions of Guer-rero and Oaxaca, Mexico, Geofısica Int. 33, 373–383.

Cerjan, C., D. Kosloff, R. Kosloff, and M. Reshef (1985). A nonreflectingboundary condition for discrete acoustic and elastic wave equations,Geophysics 50, 705–708.

Chavez, M., and R. Castro (1988). Attenuation of modified Mercalli inten-sity with distance in Mexico, Bull. Seism. Soc. Am. 78, 1875–1884.

Chavez-Garcıa, F. J. and P.-Y. Bard (1994). Site effects in Mexico Cityeight years after the September 1995 Michoacan earthquakes, SoilDyn. Earthquake Eng. 13, 229–247.

Chavez-Garcıa, F. J. and J. Cuenca (1996). Site effects in Mexico Cityurban zone: a complementary study, Soil Dyn. Earthquake Eng. 15,141–146.

Courboulex, F., M. A. Santoyo, J. F. Pacheco, and S. K. Singh (1997). The14 September 1995 (M � 7.3) Copala, Mexico, earthquake: a studyusing teleseismic, regional, and local data, Bull. Seism. Soc. Am. 87,999–1010.

Furumura, T., and B. L. N. Kennett (1998). On the nature of regional seis-mic phases III. The influence of crustal heterogeneity on the wavefieldfor subduction earthquakes: the 1985 Michoacan and 1995 Copala,Guerrero, Mexico earthquakes, Geophys. J. Int. 135, 1060–1084.

Furumura, T., and K. Koketsu (2000). Parallel 3-D simulation of groundmotion for the 1995 Kobe earthquake: The component decompositionapproach, Pure Appl. Geophys. 157, 2047–2062.

Furumura, T., and H. Takenaka (1996). 2.5-D modelling of elastic wavesusing the pseudospectral method, Geophys. J. Int. 124, 820–832.

Furumura, T., B. L. N. Kennett, and H. Takenaka (1998). Parallel 3-Dpseudospectral simulation of seismic wave propagation, Geophysics63, 279–288.

Furumura, T., K. Koketsu, and H. Takenaka (2000). A Hybrid PSM/FDMparallel simulation for large-scale 3D seismic (acoustic) wavefield,Butsuri-Tansa (J. SEGI) 53, 294–308, (in Japanese).

Furumura, T., K. Koketsu and K.-L. Wen (2002). Parallel PSM/FDM hybridsimulation of ground motions from the 1999 Chi-Chi, Taiwan, earth-quake, Pure Appl. Geophys. 159 (in press).

Garcia, D., S. K. Singh, M. Ordaz, and M. Herraiz (2001). Determinacionde una ley de atenuacion espectral para sismos intraplaca en Mexico,Extended Abstract, II Congreso Iberoamericano de Ingenieria Sis-mica, Madrid, 16–17 October.

Gonalez-Ruiz, J. R. (1986). Earthquake source mechanics and tectono-physics of the middle America subduction zone in Mexico, Ph.D.Thesis, University of California, Santa Cruz.

Hernandez, B., N. Shapiro, S. K. Singh, J. F. Pacheco, F. Cotton, M. Cam-pillo, A. Iglesias, V. Cruz, J. M. Gomez, and L. Alcantara (2001).Rupture history of September 30, 1999 intraplate earthquake of Oa-xaca, Mexico (Mw � 7.5) from inversion of strong-motion data, Geo-phys. Res. Lett. 28, 363–366.

Herrmann, R. B. (1979). S-wave generation by dislocation source: a nu-merical study, Bull. Seism. Soc. Am. 69, 1–15.

Hermann, R. B., and A. Kijko (1983). Modeling some empirical verticalcomponent Lg relations, Bull. Seism. Soc. Am. 73, 157–171.

Iglesias, A., S. K. Singh, J. F. Pacheco, and M. Ordaz (2002). A sourceand wave propagation study of the Copalillo, Mexico earthquake ofJuly 21, 2000 (Mw � 5.9): implications for seismic hazard in MexicoCity from inslab earthquakes, Bull. Seism. Soc. Am. 92, 1060–1071.

Iida, M (1999) Excitation of high-frequency surface waves with long du-ration in the Valley of Mexico, J. Geophys. Res. 104, 7329–7345.

Kawase, H., and K. Aki (1989). A study on the response of a soft basin forincident S, P, and Rayleigh waves with spectral reference to the long

duration observed in Mexico City, Bull. Seism. Soc. Am. 79, 1361–1382.

Kennett, B. L. N. (1985). On regional S. Bull. Seism. Soc. Am. 75, 1077–1086.

Kosloff, D., M. Reshef, and D. Loewenthal (1984). Elastic wave calculationby the Fourier method, Bull. Seism. Soc. Am. 74, 875–891.

Kostoglodov, V., W. Bandy, J. Domınguez, and M. Mena (1996). Gravityand seismicity over the Guerrero seismic gap, Mexico, Geophys. Res.Lett. 23, 3385–3388.

Mikumo, T., S. K. Singh, and M. Santoyo (1999). A possible stress inter-action between large thrust and normal faulting earthquakes in theMexican subduction zone, Bull. Seism. Soc. Am. 89, 1418–1427.

Ordaz, M., and S. K. Singh (1992). Source spectra and spectral attenuationof seismic waves from Mexican earthquakes, and evidence of ampli-fication in the hill zone of Mexico City, Bull. Seism. Soc. Am. 82,24–43.

Ottemoller, L., N. M. Shapiro, S. K. Shigh, and J. F. Pacheco (2002). Lat-eral variation of Lg wave propagation in southern Mexico, J. Geo-phys. Res. 10, 1029.

Pacheco, J. F., and S. K. Singh (1995). Estimation of ground motion in theValley of Mexico from normal-faulting, intermediate-depth earth-quakes in the subducted Cocos plate, Earthquake Spectra 11, 233–248.

Pardo, M., and G. Suarez (1995). Shape of subducted Rivera and Cocosplates in Southern Mexico: seismic and tectonic implications, J. Geo-phys. Res. 100, 12,357–12,373.

Quintanar, L., J. Yamamoto, and Z. Jimenez (1999). Source mechanism oftwo 1994 intermediate-depth-focus earthquakes in Guerrero, Mexico,Bull. Seism. Soc. Am. 89, 1004–1018.

Reshef, M., D. Kosloff, M. Edwards, and C. Hsiung (1988). Three-dimen-sional elastic modeling by the Fourier method, Geophysics 53, 1184–1199.

Sanchez-Sesma, F. J., S. Chavez-Perez, M. Suarez, M. A. Bravo, and L. E.Perez-Rocha (1988). On the seismic response of the Valley of Mex-ico, Earthquake Spectra 4, 569–589.

Sanchez-Sesma, F. J., L. E. Perez-Rocha, and E. Reinoso (1993). Groundmotion in Mexico City during the April 25, 1989, Guerrero earth-quake, Tectonophysics 218, 127–140.

Shapiro, N. M., M. Campillo, A. Paul, S. K. Singh, D. Jongmans, and F. J.Sanchez-Sesma (1997). Surface-wave propagation across the Mexi-can Volcanic Belt and the origin of the long-period seismic-waveamplification in the Valley of Mexico, Geophys. J. Int. 128, 151–166.

Singh, S. K., and M. Wyss (1976). Source parameters of the Orizaba earth-quake of August 28, 1973, Geofisica Int. 16, 165–184.

Singh, S. K., V. Kostoglodov, and J. F. Pacheco (2000a). Intraslab earth-quakes in the subducting oceanic plates below Mexico (abstract), Ab-stracts of the Intraslab Earthquakes Workshop, Victoria, B.C., 18–21September, 59–63.

Singh, S. K., E. Mena, and R. Castro (1988). Some aspects of source char-acteristics of the 19 September 1985 Michoacan earthquake andground motion in and near Mexico City from strong motion data, Bull.Seism. Soc. Am. 78, 451–477.

Singh, S. K., A. Mori, E. Mena, F. Kruger, and R. Kind (1990). Evidencefor anomalous body-wave radiation between 0.3 and 0.7 Hz from the1985 September 19 Michoacan, Mexico earthquake, Geophys. J. Int.101, 37–48.

Singh, S. K., M., Ordaz, L. Alcantara, N. Shapiro, V. Kostoglodov, J. F.Pacheco, S. Alcocer, C. Gutierrez, R. Quaas, T. Mikumo, and E.Ovando (2000b). The Oaxaca earthquake of September 30, 1999 (Mw

� 7.5): a normal-faulting event in the subducted Cocos Plate, Seism.Res. Lett. 71, 67–78.

Singh, S. K., M. Ordaz, J. F. Pacheco, R. Quaas, L. Alcantara, S. Alcocer,C. Gutierrez, R. Meli, and E. Ovando (1999). A preliminary reporton the Tehuacan, Mexico earthquake of June 15, 1999 (Mw � 7.0),Seism. Res. Lett. 70, 489–504.

Singh, S. K., R. Quaas, M. Ordaz, F. Mooset, D. Almora, M. Torres, and

Page 16: Regional Wave Propagation from Mexican Subduction Zone ... · Bulletin of the Seismological Society of America, Vol. 92, No. 6, pp. 2110–2125, August 2002 Regional Wave Propagation

Regional Wave Propagation from Mexican Subduction Zone Earthquakes 2125

R. Vasquez (1995). Is their truly a “Hard” rock site in the Valley ofMexico? Geophys. Res. Lett. 22, 481–484.

Singh, S. K., M. Reichle, and J. Havskov (1980). Magnitude and epicenterdeterminations of Mexican earthquakes from isoseismic maps, Geo-fısica Int. 19, 269–284.

Singh, S. K., G. Suarez, and T. Domınguez (1985). The Oaxaca, Mexico,earthquake of 1931: lithospheric normal faulting in the subductedCocos plate, Nature 317, 56–58.

Valdes, C. M., and R. P. Meyer (1996). Seismic structure between thePacific coast and Mexico City from the Petatlan earthquake (Ms �

7.6) aftershocks, Geofisica Int. 35, 377–401.Valdes, C. M., W. D. Mooney, S. K. Singh, R. P. Meyer, C. Lomnitz, J. H.

Luetgert, C. E. Helsley, B. T. R. Lewis, and M. Mena (1986). Crustalstructure of Oaxaca, Mexico, from seismic refraction measurements,Bull. Seism. Soc. Am. 76, 547–563.

Yamamoto, J., Z. Jimenez, and R. Mota (1984) El temblor de Huajuapan

de Leon, Oaxaca, Mexico del 24 de Octobre de 1980, Geofisica Int.23, 93–110.

Earthquake Research InstituteUniversity of Tokyo1-1-1 Yayoi, Bunkyo-ku113-0032, [email protected]

(T.F.)

Instituto de GeofısicaUniversidad Nacıonal Autonoma de MexicoCiudad UniversitariaCoyoacan, 04510, Mexico D.F.

(S.K.S.)

Manuscript received 23 October 2001.